Daniil Sorokin
Senior Deep Learning Engineer AI Researcher, LLM Accuracy Evaluation
Berlin, Germany
Summary
👤
Senior🎓
Top SchoolDaniil Sorokin is a Senior Deep Learning Engineer and AI researcher based in Berlin, specializing in LLM accuracy evaluation and efficient large-scale model training. With a PhD from TU Darmstadt and over a decade of applied NLP research, he has driven pre-training, tokenizer design, and early-stage evaluation for Amazon Nova and Titan models and now leads LLM health and accuracy work at NVIDIA. His background spans productionizing NLU pipelines for Alexa, weak and incremental learning, and knowledge-grounded QA, combining deep academic publication experience with hands-on deployment. He’s known for pragmatic optimizations—like custom tokenizers and training-schedule tuning—that improve throughput and model utility for coding and multilingual tasks.
11 years of coding experience
7 years of employment as a software developer
Master's Degree Applied Linguistics, Master's Degree Applied Linguistics at Saint Petersburg State UniversityTechnischen Universität DarmstadtMaster's Degree Computational Linguistics, Master's Degree Computational Linguistics at University of TübingenMachine Learning Summer School 2016 in Cadiz
English, German, Russian, French
Stackoverflow
1reputation
0reached
0answers
0questions
Github Skills (23)
pytorch10
relation-extraction10
python10
machine-learning10
extraction10
representations10
deep-learning10
gpu10
relation10
data-science9
numpy9
tensorflow9
neural-network9
gpu-acceleration9
nlp9
Programming languages (2)
Github contributions (5)
UKPLab/emnlp2017-relation-extraction
Aug 2017 - Dec 2019
Context-Aware Representations for Knowledge Base Relation Extraction
Contributions:1 release, 49 commits, 1 PR in 2 years 4 months
nlpcontext-awareextractionrepresentationsknowledge
UKPLab/emnlp2018-argmin-commonsense-knowledge
Apr 2018 - Nov 2018
Contributions:19 commits, 1 branch in 6 months
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.